netdev - RE: [PATCH net-next v4 1/2] net: libwx: support multiple RSS for every pool

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <038c01dc2775$9f4c58f0$dde50ad0$@trustnetic.com>
Date: Wed, 17 Sep 2025 09:51:44 +0800
From: Jiawen Wu <jiawenwu@...stnetic.com>
To: "'Jakub Kicinski'" <kuba@...nel.org>
Cc: <netdev@...r.kernel.org>,
	"'Andrew Lunn'" <andrew+netdev@...n.ch>,
	"'David S. Miller'" <davem@...emloft.net>,
	"'Eric Dumazet'" <edumazet@...gle.com>,
	"'Paolo Abeni'" <pabeni@...hat.com>,
	"'Simon Horman'" <horms@...nel.org>,
	"'Alexander Lobakin'" <aleksander.lobakin@...el.com>,
	"'Mengyuan Lou'" <mengyuanlou@...-swift.com>
Subject: RE: [PATCH net-next v4 1/2] net: libwx: support multiple RSS for every pool

On Tue, Sep 16, 2025 9:02 AM, Jakub Kicinski wrote:
> On Fri, 12 Sep 2025 14:23:56 +0800 Jiawen Wu wrote:
> > Subject: [PATCH net-next v4 1/2] net: libwx: support multiple RSS for every pool
> 
> "support multiple RSS" needs an object. Multiple RSS keys? Multiple
> contexts? Multiple tables?

All of these are multiple. Each pool has a different RSS scheme.

> 
> > -static void wx_store_reta(struct wx *wx)
> > +u32 wx_rss_indir_tbl_entries(struct wx *wx)
> >  {
> > +	if (test_bit(WX_FLAG_SRIOV_ENABLED, wx->flags))
> > +		return 64;
> > +	else
> > +		return 128;
> > +}
> 
> Is WX_FLAG_SRIOV_ENABLED set only when VFs are created?

Yes.

> What if the user set a table with 128 entries?
> The RSS table can't shrink once intentionally set to a specific size.

Deleting VFs will reset these configurations.

> 
> > +void wx_store_reta(struct wx *wx)
> > +{
> > +	u32 reta_entries = wx_rss_indir_tbl_entries(wx);
> >  	u8 *indir_tbl = wx->rss_indir_tbl;
> >  	u32 reta = 0;
> >  	u32 i;
> > @@ -2007,36 +2016,55 @@ static void wx_store_reta(struct wx *wx)
> >  	/* Fill out the redirection table as follows:
> >  	 *  - 8 bit wide entries containing 4 bit RSS index
> >  	 */
> > -	for (i = 0; i < WX_MAX_RETA_ENTRIES; i++) {
> > +	for (i = 0; i < reta_entries; i++) {
> >  		reta |= indir_tbl[i] << (i & 0x3) * 8;
> >  		if ((i & 3) == 3) {
> > -			wr32(wx, WX_RDB_RSSTBL(i >> 2), reta);
> > +			if (test_bit(WX_FLAG_SRIOV_ENABLED, wx->flags) &&
> > +			    test_bit(WX_FLAG_MULTI_64_FUNC, wx->flags))
> > +				wr32(wx, WX_RDB_VMRSSTBL(i >> 2, wx->num_vfs), reta);
> 
> Do we need to reprogram the RSS when number of VFs change, now?

Yes. We program RSS in the device open function, which is called then
the number of VFs changed.

> 
> > +			else
> > +				wr32(wx, WX_RDB_RSSTBL(i >> 2), reta);
> >  			reta = 0;
> >  		}
> >  	}
> >  }
> >
> > +void wx_store_rsskey(struct wx *wx)
> > +{
> > +	u32 random_key_size = WX_RSS_KEY_SIZE / 4;
> 
> They key is just initialized to a random value, it doesn't have to be
> random. Just "key_size" is better.
> 
> > +	u32 i;
> > +
> > +	if (test_bit(WX_FLAG_SRIOV_ENABLED, wx->flags) &&
> > +	    test_bit(WX_FLAG_MULTI_64_FUNC, wx->flags)) {
> > +		for (i = 0; i < random_key_size; i++)
> > +			wr32(wx, WX_RDB_VMRSSRK(i, wx->num_vfs),
> > +			     *(wx->rss_key + i));
> 
> Prefer normal array indexing:
> 
> 			     wx->rss_key[i]
> 
> > +	} else {
> > +		for (i = 0; i < random_key_size; i++)
> > +			wr32(wx, WX_RDB_RSSRK(i), wx->rss_key[i]);
> > +	}
> > +}
> 
> > -	u32 rss_field = 0;
> 
> completely unclear to me why moving rss_field to struct wx is part of
> this patch. It looks unrelated / prep for the next patch.

I'll split it, thanks.